Linus Torvalds Improves Futex Code To Improve User-Space Accesses

[phoronix]

Phoronix: Linus Torvalds Improves Futex Code To Improve User-Space Accesses

In between managing all of the pull requests being submitted during this two week long merge window for the Linux 6.13 kernel, Linus Torvalds has merged some of his own code this cycle...

https://www.phoronix.com/news/Linux-6.13-Torvalds-Futex

[nuetzel]

NTSYNC, please.

[LuukD]

Some articles get hundreds of comments. Drama I suppose. But this one is nearly void of any, while improving futexes is big!
Futexes make up the foundation of synchronization primitives for many languages. Languages like C/C++ using pthreading or indirectly through, MPI or OpenMP, Rust std, Java HotSpot, the CPython interpreter all rely on futexes on Linux.

Given that our software is getting increasingly multi threaded, further reducing contention for synchronization primitives is increasingly important.

I am amazed that the kernel developers are still able to squeeze more performance from futexes since they - as part of the kernels hot paths - get a lot of attention. User space multi-threaded performance relies on these and its worthwhile to invest here. Companies and institutions likely care because they want to make better use of their costly many-core machines. Yet they still found another ~2%. If only the misbehaving architectures could change their ways.

[User29]

I am amazed that the kernel developers are still able to squeeze more performance from futexes since they - as part of the kernels hot paths - get a lot of attention.

this is exactly what I was thinking about, without checking the futex-related commit history. how come that there are better and better algorithms are coming up. this is really amazing.